Ant Group’s Robbyant Extends World Model Generation to Hour-Long Real-Time Output
Event summary
- Robbyant, an Ant Group subsidiary, released LingBot-World 2.0 (Infinity), enabling hour-long real-time generation in world models.
- The model supports 720p/60fps high-definition output and introduces a native agent mechanism for sustainable interactivity.
- LingBot-World 2.0 features a Causal Pretraining Paradigm with MoBA (Mask of Bidirectional Attention) to eliminate long-horizon generation errors.
- The model is available on Reactor, GitHub, and Hugging Face under an open-source license.
The big picture
Robbyant’s LingBot-World 2.0 represents a significant leap in world model technology, pushing the boundaries of real-time generation and interactivity. This advancement aligns with the growing trend of open-source contributions in AI, potentially reshaping how developers approach embodied intelligence applications.
What we're watching
- Adoption Pace
- How quickly developers and enterprises integrate LingBot-World 2.0 into real-world applications.
- Competitive Response
- Whether rivals in the world model space accelerate their own advancements to match Robbyant’s capabilities.
- Open-Source Impact
- The extent to which open-sourcing LingBot-World 2.0 fosters broader innovation and collaboration in embodied AI.
Related topics